guangerrr/UrbanGovQA

UrbanGovQA是一个用于在城市治理领域训练和对齐大型语言模型的数据集。它旨在支持监督微调（SFT）和直接偏好优化（DPO），并明确区分了两种互补的任务类型：理论（T）和应用（A）。理论任务涉及概念理解、政策机制、制度分析和比较推理，而应用任务则涉及基于场景的推理、政策实施、决策支持和可操作分析。数据集通过权威的城市治理文献和严格的过滤规则生成，旨在提供高质量且一致的监督和偏好数据。

UrbanGovQA is a dataset for training and aligning large language models in the domain of urban governance. It is designed to support both Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), and explicitly distinguishes between two complementary task types: Theory (T) and Applied (A). Theory tasks involve conceptual understanding, policy mechanisms, institutional analysis, and comparative reasoning, while Applied tasks involve scenario-based reasoning, policy implementation, decision support, and actionable analysis. The dataset is generated using authoritative urban governance literature and strict filtering rules, aiming to provide high-quality and consistent supervised and preference data.